Why Susie Sells Seashells by the Seashore: Implicit Egotism and Major Life Decisions

ATTITUDES AND SOCIAL COGNITION

Brett W. Pelham, Matthew C. Mirenberg, and John T. Jones State University of New York at Buffalo

Because most people possess positive associations about themselves, most people prefer things that are connected to the self (e.g., the letters in one’s name). The authors refer to such preferences as implicit egotism. Ten studies assessed the role of implicit egotism in 2 major life decisions: where people choose to live and what people choose to do for a living. Studies 1–5 showed that people are disproportionately likely to live in places whose names resemble their own first or last names (e.g., people named Louis are disproportionately likely to live in St. Louis). Study 6 extended this finding to birthday number preferences. People were disproportionately likely to live in cities whose names began with their birthday numbers (e.g., Two Harbors, MN). Studies 7–10 suggested that people disproportionately choose careers whose labels resemble their names (e.g., people named Dennis or Denise are overrepresented among dentists). Implicit egotism appears to influence major life decisions. This idea stands in sharp contrast to many models of rational choice and attests to the importance of understanding implicit beliefs.

What role do people’s thoughts and feelings about themselves important role in major life decisions. For example, only a handful play in their important day-to-day decisions and behaviors? Con- of studies have examined whether self-regulation processes influ- temporary research on the self-concept suggests many answers to ence people’s choice of relationship partners. Moreover, for prac- this question. For example, the desire to maintain positive feelings tical reasons, most of these studies have examined attraction to about the self appears to influence things as diverse as whether strangers in the laboratory rather than attraction to long-term people derogate those who criticize their governments (Pyszczyn- relationship partners (Huston & Levinger, 1978; but cf. Murray, ski, Greenberg, & Solomon, 1997, 1999), whether people sabotage Holmes, & Griffin, 1996; Swann, Hixon, & de la Ronde, 1992). the performance of others when playing a game (Tesser & Smith, We thus know relatively little about whether self-evaluations or 1981), what people find rational (Kunda, 1990), and what people self-concept motives influence important decisions such as where find humorous (Wills, 1981). Like people’s social beliefs and people choose to live and what people choose to do for a living. behavior, people’s beliefs about themselves are also influenced Is there any reason to believe that self-evaluation shapes impor- by the desire to view the self favorably. Most people have tant life decisions? Consider the decisions of (a) choosing a city or mostly favorable beliefs about themselves (see Crocker & Ma- state in which to live and (b) choosing a career. At first blush, it jor, 1989; Greenberg & Pyszczynski, 1985; Miller & Ross, might seem that these decisions are largely independent. That is, 1975; Paulhus & Levitt, 1987; Taylor & Brown, 1988; but cf. there would appear to be no unifying psychological thread that Kruger, 1999). connects a person’s taste in cities and career tracks. This does not In short, a great deal of evidence suggests that the motivation to mean that these important decisions are completely unrelated. feel good about the self plays a role in a wide variety of important Consider the dentist who is offered jobs in Milwaukee and in social behaviors. To our knowledge, however, very little research Phoenix. She might choose the job in Phoenix—either because it on the self-concept addresses whether self-evaluation plays an offers her greater professional rewards or because it offers her the promise of milder winters. Such obvious connections aside, however, it might be asking quite a bit of a single self-evaluative mechanism to predict people’s taste in both hometowns and Brett W. Pelham, Matthew C. Mirenberg, and John T. Jones, Depart- occupations. ment of Psychology, State University of New York at Buffalo. We thank Mauricio Carvallo, Joanne Davila, Tracy Dehart, Curtis Hardin, Sandra Murray, Shirley Owens, Mike Ross, and the participants Implicit Egotism of the 2000 Ontario Symposium for their insightful comments on this research. Or would it? Research in implicit social cognition suggests that Correspondence concerning this article should be addressed to Brett W. many mundane judgments and behaviors are influenced by potent Pelham, State University of New York, University at Buffalo, Park Hall— and pervasive unconscious motives. For example, Greenwald and Department of Psychology, Buffalo, New York 14260-4110. E-mail: Banaji (1995) argued that the desire to feel good about the self [email protected] permeates a wide range of social judgments. As an example,

469 470 PELHAM, MIRENBERG, AND JONES students’ evaluations of instructors’ teaching skills tend to be more mere exposure. Instead, like the mere ownership effect, the favorable when instructors give more lenient grades. Presumably, name letter effect appears to qualify as a form of implicit this is true because instructors who give students high grades egotism (Greenwald & Banaji, 1995; Nuttin, 1987). The essen- satisfy students’ desires for favorable feedback. That is, uncon- tial idea behind implicit egotism is that people should prefer scious self-enhancement seems to bias students’ evaluations of people, places, and things that they associate (unconsciously) their instructors’ teaching skills. As noted previously, researchers with the self. have long assumed that people have a basic desire to feel good Although examples of unconscious self-enhancement have been about themselves (Allport, 1961; James, 1890/1950). What sets documented for a variety of judgments and decisions, virtually all recent research on self-enhancement apart from a great deal of of the past research on unconscious self-enhancement simply earlier research is the contemporary assumption that many self- documents the existence of positive associations to the self (but see enhancing social cognitions occur automatically or unconsciously Greenwald & Banaji, 1995, for some exceptions). For instance, (Greenwald & Banaji, 1995; Hetts, Sakuma, & Pelham, 1999; whereas existing research leaves little doubt that people like the Jones, Pelham, Mirenberg, & Hetts, 2002; Paulhus & Levitt, 1987; letters in their own names, we know of no previous research Pelham & Hetts, 1999; Perdue, Dovidio, Gurtman, & Tyler, 1990; examining behavioral consequences of the name letter effect. To Pyszczynski et al., 1997, 1999). our knowledge, the closest thing to this is Miller, Downs, and Research on unconscious self-enhancement—or what we call Prentice’s (1998) finding that people are nicer than usual to strang- implicit egotism—suggests that people’s positive automatic asso- ers who happen to share their birthdays. In particular, Miller et ciations about themselves may influence their feelings about almost anything that people associate with the self. For example, al.’s participants played more cooperatively than usual with a research on the mere ownership effect shows that giving people bogus interaction partner (in a prisoner’s dilemma game) when objects such as pens or keychains causes people to evaluate these they thought they and their partner had the same birthday (see also object more favorably than they would otherwise (Beggan, 1992; Finch & Cialdini, 1989). If people truly prefer the specific letters Kahneman, Knetsch, & Thaler, 1990; Van Boven, Dunning, & that appear in their own names, then we might expect people to Loewenstein, 2000). If people instantly acquire positive feelings prefer physical or social objects that either contain or are repre- about objects once these objects become part of the self, it stands sented by these same letters. For example, consider the case of a to reason that people should develop deep and abiding affections hypothetical person named Steve Spencer. If the name letter effect for objects that are chronically associated with the self. In support has implications for important social behaviors, it might not be too of this idea, research on the name letter effect (Kitayama & far-fetched to expect Steve to prefer to vacation in Sarasota rather Karasawa, 1997; Nuttin, 1985, 1987) shows that people like the than Miami, to choose to study stereotyping rather than attitudes, letters that appear in their own names quite a bit more than they or to find himself collaborating with colleagues named Steven Fine like letters that do not. This effect occurs for all of the letters in (or Clyde Steele) rather than Russ Lazio. Although Steve would people’s names, but it is particularly pronounced for people’s first probably be unaware of the affective basis of his judgment, some- and last initials. This presumably unconscious self-enhancing bias thing about names such as Sarasota, stereotyping, and Steele has been documented in at least 14 countries (e.g., Greece, Japan, should just feel good (Greenwald & Banaji, 1995; Lewicki, 1985; Spain, the United States). Nisbett & Wilson, 1977; Staats & Staats, 1957, 1958). Although a high level of exposure to the letters that occur in In this article, we present evidence from a wide range of archival one’s own name probably plays a role in the development of the studies that examined two different behavioral consequences of name letter effect (see Zajonc, 1968), it seems unlikely that the implicit egotism: (a) choosing a city or state in which to live and name letter effect is determined exclusively by mere exposure (b) choosing a career. Because the name letter effect lends itself (Nuttin, 1987). Consider a recent study of unmarried Japanese particularly well to archival research using public records (e.g., college students conducted by Kitayama and Karasawa (1997). telephone directories, memberships in professional organizations), These researchers observed the strongest evidence for the name we focus primarily on behavioral consequences of the name letter letter effect for the first letter of women’s first names and for the effect in this research. However, we assume that many other forms first letter of men’s last names. They explained this gender differ- of implicit egotism (e.g., the preferences people have for their ence by noting that there is a very strong expectation that Japanese birthday numbers, implicit in-group bias) also have important women will change their family names when they marry. Thus, consequences for day-to-day decisions. In the present studies, we although there is no reason to assume that men and women differ in their level of exposure to their first versus last initials, women made use of public records to identify people by their names or appear to have more positive feelings about the specific name initials. We then examined whether people with specific kinds of letters that they know they will hold onto for their entire lives. names or initials (e.g., people named Dennis) gravitated toward Kitayama and Karasawa also found that people preferred numbers specific places in which to live (e.g., Denver) or specific occupa- that were in their birthdays to numbers that were not. Moreover, in tions (e.g., dentistry). Because past research suggested that the the case of the birthday numbers that do not overlap with birthday name letter effect is stronger for men’s last names and for wom- months (13–31), the magnitude of the evaluative bias associated en’s first names, we took this into account wherever it was perti- with birthday numbers was slightly larger than the magnitude of nent. On the whole, however, we hoped to show that the name the name letter effect. Because people are exposed to their birthday letter effect has important behavioral consequences for men and numbers much less often than they are exposed to their names, this women alike. In our first study, we tested the idea that the positive further suggests that people’s elevated liking for objects that are associations people have for the letters in their names influence the associated with the self is not simply an affective consequence of city in which people choose to live. IMPLICIT EGOTISM IN MAJOR LIFE DECISIONS 471

Study 1 highly significant, ␹2(1) ϭ 38.25, p Ͻ .001. As shown in the lower half of Table 1, the results for men were also highly consistent with Method predictions. Just as Jacksonville seems to have been disproportion- In Study 1, we analyzed social security death index (SSDI) records ately inhabited by Jacks, Philadelphia seems to have been dispro- 2 available at a genealogical Web site (http://www.Ancestry.com). The portionately inhabited by Philips, ␹ (1) ϭ 202.05, p Ͻ .001. searchable database available at this Web site is frequently updated and Because all of the names used in Study 1 are common European was generated from U.S. government records in May of 2001. These American names, it is difficult to attribute these findings to an records include the more than 66 million deceased Americans who pos- ethnic confound. It is even more difficult to attribute these findings sessed social security numbers at their time of deaths. These SSDI records to an age confound. In fact, to begin with the women, an exami- and their associated search engine are ideal for the present purposes nation of the birth dates of all of the decedents of Virginia Beach because they (a) are highly comprehensive, (b) allow searches based on and Milwaukee indicated that the average resident of Milwaukee either first or last names, and (c) allow searches based on a person’s last was noticeably older than the average resident of Virginia Beach known city or state of residence prior to death. Although it is not possible to download these data, the search engine that is a part of this Web site (as indicated by a greater ratio of people born in 1990 versus places no restrictions on the number of hits it returns (e.g., it makes it 1950). The small difference in the ages of the two female names possible to determine exactly how many people named Gladys or Ezekiel used in Study 1 worked in opposition to this city–age confound. have died in the United States since the beginning of the social security Specifically, the name Mildred is slightly younger than the name system). Virginia. The same was true for the male names. Whereas the In Study 1, we first identified the 40 largest cities in the United States. decedents of Philadelphia appear to have been somewhat older, on We then consulted the 1990 census to identify all of the common (top 100) average, than the decedents of Jacksonville, Philip turns out to be male and female first names that shared a minimum of their first three a slightly younger name than Jack. letters with any of these city names. Because the popularity of different Although the results of Study 1 suggest that implicit egotism first names varies with age, we selected the two qualifying European may play a role in people’s choice of a hometown, these results are American female names that we could match most closely for age (based on the relative proportion of deceased people born in 1900 and 1950 who open to a highly plausible alternate explanation. Although it is had these first names). We then repeated this procedure to produce the two possible that people gravitate toward places that remind them of most closely age-matched male names. The resulting name–city combina- themselves, it is also possible that the places in which people live tions for women were Mildred–Milwaukee and Virginia–Virginia Beach. serve as primes that influence the names that parents give to their The resulting combinations for men were Jack–Jacksonville and Philip– children. To new parents living in Georgia, for example, baby Philadelphia. To generate the data for Study 1, we created two separate names such as George or Georgia may simply be more accessible 2 ϫ 2 name–city tables, one for women and one for men. We expected that than more attractive alternatives such as Brett, John, or Matthew. the women and men whose first names resembled the name of a specific Although this priming effect would be interesting in its own right, city would be over-represented among the deceased residents of that it has little to do with implicit egotism. In Study 2, we attempted specific city. to rule out this alternate explanation by focusing on surnames rather than first names. Because laboratory research has shown that Results and Discussion surname letter liking is typically lower than first name letter liking, The results for women appear in the upper half of Table 1. As we suspected that surname effects might be smaller than first name shown in Table 1, women named Mildred and Virginia did seem effects. In light of this expectation, we sampled U.S. states rather to gravitate toward cities that resembled their first names. The than large cities. If people with surnames that resemble specific association between name and place of residence for women was state names disproportionately populate these states, we can be relatively confident that the owners of these surnames (or, at a minimum, at least one of their ancestors) chose to live in these specific states. Table 1 Likelihood of Living in a City as a Function of One’s First Name (Study 1) Study 2

First name: Women Method

City Mildred Virginia Total Study 2 was an expanded version of Study 1. However, in an effort to Milwaukee 865 (806) 544 (603) 1409 generalize our findings to a wide range of surnames, we conducted sur- Virginia Beach 230 (289) 275 (216) 505 name searches based on the first few consecutive letters of each of the state Total 1095 819 1914 names we sampled (rather than searching for one specific surname per state). We began by identifying an electronic telephone directory that was First name: Men ideal for the present purposes, namely the WorldPages directory (http:// www.Worldpages.com). This directory will return as many as 2,000 hits in Jack Philip Total a specific U.S. or Canadian city, state, or province. We chose eight large U.S. states on the basis of population. Specifically, on the basis of 1990 Jacksonville 436 (288) 111 (259) 547 Philadelphia 968 (1116) 1153 (1005) 2121 census data, we identified the eight most populous U.S. states that had Total 1404 1264 2668 one-word names. (We eliminated states with two-word names because, with very few exceptions, people do not have two-word surnames. In Note. Expected values (rounded to the nearest whole number) appear in addition, the first word in most two-word state names is more peripheral to parentheses. the state name than is the second word.) We selected the initial letter strings 472 PELHAM, MIRENBERG, AND JONES for these searches by beginning with the first four letters of each state name to be Latino, it is extremely difficult to attribute the large effects in our list. When a four-letter string (e.g., Geor for Georgia) yielded too observed for these two states to a confound involving Latino many hits to be listed by the search engine, we added the next consecutive versus European names. letter(s) in the state name to narrow the search. When a four-letter string In short, despite the risks associated with drawing causal infer- yielded no or very few hits (e.g., Ohio), we deleted letters from the end of ences from passive observational findings, we suggest that the best the string, one at a time, to maximize number of hits realized. For example, explanation for the findings of Studies 1–2 is that people are we used Ohi (which fit within the 2,000 hit limit for all eight states) rather than Oh (which often exceeded the limit). Thus, in the case of every state attracted to places that resemble their own names. Furthermore, in our list, we generated an initial surname letter string set that (a) matched although creative readers might be able to generate additional each state name as well as possible and (b) maximized search size on the criticisms of these findings, alternate explanations based on eth- basis of the 2,000 hit constraint. nicity seem relatively implausible, and alternate explanations The only exception to this procedure occurred for Pennsylvania. Penn based on reverse causality (i.e., priming) no longer seem plausible yielded too many hits, and Penns yielded almost none. Our solution was to at all. add y rather than s to Penn, yielding Penny. The two most obvious alternate Because these findings challenge traditional assumptions about solutions to this problem (either dropping Pennsylvania altogether or how people make major life decisions, critics may still wish to adding the final letter a rather than the next consecutive letter y to Penn) reserve judgment. One could argue, for example, that many of the yielded findings slightly stronger than those reported herein. Because this surnames that were likely to have been generated in Study 2 were directory occasionally yielded hits that appeared to be clubs, institutions or businesses (e.g., Florida, Gators) rather than people (e.g., Florian, David) rather unusual. The results of Study 1 notwithstanding, are the we eliminated all such noisy hits within each state. Incidentally, the SSDI examples of implicit egotism documented in Study 2 limited to search engine used in Study 1 was not appropriate for Study 2 because it highly unusual names? Do such findings generalize to people’s only allows searches based on exact names (rather than the first few letters choice of cities rather than states? Study 3 was designed to address of a name). these questions.

Results and Discussion Study 3 The results of Study 2 are summarized in Table 2. For five of the eight state–name pairs, the results were weakly to strongly in the Method predicted direction; for one of the pairs, the observed frequencies Study 3 was patterned directly after Study 2. Thus, Study 3 also focused were very close to the expected frequencies; and for two pairs on surnames and also made use of the WorldPages electronic telephone (Pennsylvania and Ohio), there were modest reversals. The find- directory. In Study 3, however, we focused on the eight largest Canadian ings for Ohio are likely a product of sampling error. This may be cities rather than the eight largest U.S. states. More specifically, we true of Pennsylvania as well. For instance, if one conducts a search identified the eight largest predominantly English-speaking metropolitan based on names beginning with Penna, the results for Pennsylva- areas in Canada. We excluded Montreal and Quebec to avoid an English– nia look much more reasonable (112.8 expected vs. 135 observed French ethnic name confound (e.g., if a disproportionate number of French hits). Regardless of one’s preferred interpretation for any reversals, names began with Mont, including Montreal in our list could falsely inflate the overall name–state matching effect (with the reversals aver- our findings). Precisely the same design and same sampling procedure used in Study 2 yielded eight sets of 3- or 4-letter surname strings, one string aged in) was highly significant. The expected number of surname– corresponding to the first 3–4 letters of each city in the list. Needless to state matches was 1,584.4 (16.64%), and the observed number of say, initial strings such as Edm (Edmonton), Ham (Hamilton), and Lon ␹2 ϭ Ͻ matches was 1,890 (19.86%), (1) 70.71, p .001. It is worth (London) yielded surnames that were quite a bit more common than those noting that the states we sampled in Study 2 provide some excel- typically generated in Study 2 (see Table 3 for the complete set of cities lent controls for potential ethnic confounds. For example, because and letter strings). As in Study 2, we deleted any hits that did not appear exactly 32% of the residents of both Texas and California happen to refer to people. Careful readers will notice that the surface differences in

Table 2 Likelihood of Living in a Specific State as a Function of One’s Surname (Study 2)

Surname initial letter string

State Cali Texa Flori Illi Penny Ohi Michi Georgi Total

California 929 (760) 10 357 155 299 66 153 264 2,233 Texas 350 34 (7) 170 68 451 31 65 76 1,245 Florida 737 3 471 (358) 95 293 18 68 218 1,903 Illinois 328 0 172 67 (63) 100 21 52 107 847 Pennsylvania 463 4 266 201 265 (302) 3 118 151 1,471 Ohio 188 0 130 40 264 2 (12) 16 110 750 Michigan 145 2 182 67 147 8 87 (44) 74 712 Georgia 98 1 42 20 135 3 24 35 (39) 358 Total 3,283 54 1,790 713 1,954 152 583 1,035 9,519

Note. States are listed in order of population. For the diagonal frequencies, in boldface type, expected values (rounded to the nearest whole number) appear in parentheses. IMPLICIT EGOTISM IN MAJOR LIFE DECISIONS 473 the sampling procedures of Studies 2 and 3 merely reflect the fact that time of their demise but also the state in which they lived when Study 3 focused on cities rather than states. Because city populations are they were issued their social security cards. noticeably smaller than state populations, the same procedure used in Because the large majority of deceased Americans obtained Study 2 typically generated three-letter rather than four-letter strings in social security cards as adults, it is important to note that SSDI Study 3 (making Study 3 more conservative than Study 2). records do not typically indicate where people were born (the social security system did not exist until 1935). This means that Results and Discussion SSDI records cannot typically distinguish a person who moved to a state as an adult from a person who was born in that state. Like their American counterparts, these Canadian residents However, they can be used to document moves among those who tended to reside in places that resembled their surnames. As happened to move after receiving their social security cards. In illustrated in Table 3, this was true for seven of the eight name– other words, studying people who changed residences as adults city pairs. The only exception to the surname–city matching rule constitutes a very conservative test of our hypothesis. With this in occurred for the name–city pair Cal–Calgary. Perhaps a few of mind, we attempted to identify a very large sample of participants these Calgarians went missing after visiting American relatives in in Study 4. We did so by focusing on relatively common first California (see Table 2, column 1). Alternately, perhaps it was names and looking at the residents of U.S. states rather than cities. satisfying enough for them merely to live elsewhere in Canada. Nonetheless, for the sample as a whole (Calgarians included), the Study 4 overall surname–city name matching effect was highly significant 2 (2,497 observed versus 2,178 expected matches), ␹ (1) ϭ 52.99, Method p Ͻ .001. Thus, even when it comes to the more run-of-the-mill surnames that were typically generated in Study 3 (e.g., Edmunds, In Study 4 we focused on states in the Southeastern U.S. This sampling London, Winters), people seem to prefer places that resemble their strategy was purely pragmatic. Relative to states in all other regions of the surnames. Although the absolute magnitude of this matching effect U.S., Southeastern states are disproportionately likely to have names that was not very large (about 15% above the chance value), this strongly resemble common person names (e.g., Georgia was named after King George). After identifying all Southeastern states with one-word seemingly small effect is not negligible. For example, this effect names, we consulted census records to identify all of the common male and size is about three times the effect size that corresponds to a female names that shared a minimum of their first 3 letters with these casino’s long-run advantage in a game of roulette. It thus seems a Southeastern state names. When more than one name matched a state safe bet that people are attracted to places whose names resemble name, we chose the most common name. This procedure yielded four their own first or last names (see Prentice & Miller, 1992). name–state pairs for both women and men (e.g., Florence–Florida, Although Studies 1–3 suggest that implicit egotism plays an George–Georgia; see Tables 4 and 5). Because we sampled multiple names important role in where people choose to live, critics have noted for both women and men, we were unable to match names directly for age that Studies 2 and 3 do not provide irrefutable evidence that people as we had done in Study 1. However, the very large sample size of Study 4 actually moved to the states or cities that resemble their surnames. allowed us to control for age directly in supplemental analyses to be Unless people routinely change their surnames to match the names described below. In addition, the large sample size available for Study 4 allowed us to perform two separate sets of analyses: (a) preliminary of the places in which they live, we do not consider this is a very analyses based simply on where people lived at the time of their deaths and compelling criticism. Nonetheless, it should go without saying that (b) more focused analyses based on the subset of people who moved to the it would be nice to have direct evidence regarding people’s move- states in our sample as adults. To determine the number of people who ment from one city or state to another. In Study 4, we attempted to moved to a specific state as adults, we simply subtracted (a) the total gather such evidence by taking advantage of the fact that SSDI number of people with a specific first name who received their social records include not only the state in which decedents lived at the security card in that state and also died while living in that state from (b)

Table 3 Likelihood of Living in a Specific Canadian City as a Function of One’s Surname (Study 3)

Surname initial letter string

City Tor Vanc Ott Edm Cal Win Ham Lon Total

Toronto 836 (611) 61 178 165 1646 1116 1923 914 6,839 Vancouver 114 21 (19) 38 48 316 322 467 192 1,518 Ottawa 104 24 52 (50) 37 303 240 553 217 1,530 Edmonton 111 23 85 54 (48) 377 361 576 193 1,780 Calgary 177 47 86 82 409 (528) 535 904 304 2,544 Winnipeg 150 20 82 45 259 338 (287) 508 157 1,559 Hamilton 98 22 45 23 238 236 527 (434) 176 1,365 London 81 22 50 46 328 289 480 260 (201) 1,556 Total 1,671 240 616 500 3,876 3,437 5,938 2,413 18,691

Note. Cities are listed in order of population. For the diagonal frequencies, in boldface type, expected values (rounded to the nearest whole number) appear in parentheses. 474 PELHAM, MIRENBERG, AND JONES the total number of people with a specific first name who simply died while women and men when we compared people who had different first living in that state (regardless of where they were born). As a concrete names but the same surnames, both ps Ͻ .001. Although this example, if (a) 100 men named George received their social security cards sampling procedure does a good job of eliminating many ethnic in Georgia and also died while living in Georgia and (b) a total of 150 men confounds (e.g., it largely eliminates people of French, German, or named George died while living in Georgia, it seems like a safe bet (barring Visigothic ancestry), it does not eliminate every possible ethnic errors in social security records) that 50 men named George moved to Georgia. Needless to say, the strongest evidence for implicit egotism would confound. In particular, many African Americans (e.g., Venus be evidence that people actually moved to states resembling their first Williams, James Brown) have one of these five highly common names. surnames. On the other hand, unless (a) some of these first names are more common than others among Blacks as opposed to Whites, Results and Discussion (b) some of these Southeastern states have noticeably larger Black populations than others, and (c) these demographic differences Preliminary results. The preliminary results of Study 4 were happen to coincide perfectly with the forename–state pairs we used highly similar to the results of Studies 1–3. First of all, as sug- in Study 4, it is not easy to attribute these findings to any obvious gested by the frequencies in the diagonals of Table 4, women ethnic confounds. Nonetheless, to address this possibility more disproportionately inhabited states that resembled their first names. directly, we conducted a number of auxiliary analyses. To briefly In fact, this was strongly true for all four of the name–state pairs. summarize these findings, the results of Study 4 are still extremely Whereas the total number of expected (chance) name–state robust when one only compares states with very similar propor- matches was 19,859, the observed number was 28,530, or about tions of Black and White residents (e.g., Georgia and Louisiana). 44% more than the chance value, ␹2(1) ϭ 9,375.74, p Ͻ .001. The Finally, these results appear to apply to the living and breathing as results for men are summarized in Table 5. In the case of men, the well as the dearly departed. In two different follow-up studies in results were weaker but still highly significant. On the whole, men which we made use of online electronic phone books (e.g., World- were about 26% more likely to live in states resembling their Pages.com) rather than SSDI records, we observed clear replica- names than they should have been by chance, ␹2(1) ϭ 2,862.00, tions of the name–state matching effect—whether we sampled p Ͻ .001. Thus, Study 4 provided a large scale, highly systematic replication of the findings of Studies 1–3. people with any surname at all or people with any of the five most Some readers may worry that confounds involving age or eth- common U.S. surnames. nicity could be responsible for the findings of Study 4. For in- Results for interstate immigrants. Preliminary analyses clearly stance, if the average resident of Florida is older than the average showed that people disproportionately inhabited states that resem- resident of Georgia, and if women named Florence tend to be older bled their names. But did these residents actually move to these than women named Georgia, this could potentially account for the states? More focused analyses indicated that they did. The results findings of Study 4. We addressed this age confound by limiting for women who migrated from one state to another after receiving the analyses of Study 4 to women and men born during exactly the their social security cards are summarized in Table 6. For all four same year. As a representative example, in one such analysis, we name–state pairs, the findings strongly supported our predictions. focused exclusively on people born in 1920. Despite a dramatic On average, women were about 18% more likely to move to states reduction in sample size, the name–state matching effect remained resembling their first names than they should have been based on highly significant for both women and men, both ps Ͻ .001. Thus, chance, ␹2(1) ϭ 1,037.57, p Ͻ .001. Furthermore, for the two an age confound cannot explain these findings. We also conducted states that happened to match women’s names perfectly (Georgia several analyses to address potential ethnic confounds. For exam- and Virginia), this effect was noticeably larger than usual (36% ple, we conducted five separate replications in which we limited higher than the chance values). The results for men are summa- our searches to people who all had the same European American rized in Table 7. As was the case in the preliminary analyses, the surname—more specifically, one of the five most common Amer- focused results for men were weaker than the results for women. ican surnames (Smith, Johnson, Williams, Jones, or Brown). We Findings were weakly to moderately supportive for three of the continued to observe robust name–state matching effects for both four name–state pairs. On average, men’s migration rates to cities

Table 4 Likelihood of Living in a Specific Southeastern State as a Function of One’s First Name (Study 4: Preliminary Results for Women)

First name

State Florence Georgia Louise Virginia Total

Florida 13,145 (9,641) 1,920 8,820 8,822 32,707 Georgia 2,591 2,202 (1,103) 5,335 2,985 13,113 Louisiana 2,646 926 4,303 (3,175) 2,054 9,929 Virginia 3,861 1,298 5,671 8,880 (5,940) 19,710 Total 22,243 6,346 24,129 22,741 75,459

Note. For the diagonal frequencies, in boldface type, expected values (rounded to the nearest whole number) appear in parentheses. IMPLICIT EGOTISM IN MAJOR LIFE DECISIONS 475

Table 5 Likelihood of Living in a Specific Southeastern State as a Function of One’s First Name (Study 4: Preliminary Results for Men)

First name

State George Kenneth Louis Virgil Total

Georgia 13,697 (12,261) 1,477 1,642 855 17,671 Kentucky 11,390 2,092 (1,636) 2,214 1,736 17,432 Louisiana 9,100 1,045 5,775 (2,665) 397 16,317 Virginia 16,629 2,261 2,332 597 (1,068) 21,819 Total 50,816 6,875 11,963 3,585 73,239

Note. For the diagonal frequencies, in boldface type, expected values (rounded to the nearest whole number) appear in parentheses. resembling their names exceeded the chance value by only about conducted several systematic pilot studies suggesting that, in ad- 4%. Nonetheless, because of the large sample size, this small dition to gender, the distinctiveness or statistical uniqueness of overall name–city migration effect for men was still highly sig- people’s names may also moderate the strength of behavioral name nificant, ␹2(1) ϭ 19.93, p Ͻ .001. As noted earlier, this test for letter matching effects. For example, to eliminate any variation in adult migration is an extremely conservative test. Presumably, how closely people’s first names matched state names (e.g., the many of the women and men in the preliminary analyses also name Georgia matches Georgia better than the name Florence migrated to the states in which they died (but did so before matches Florida), we identified the 8 most common first names receiving their social security cards). (other than Georgia and Virginia) that matched state names exactly. These eight names (generated based on their population Likely Moderators of Implicit Egotism frequencies) ranged from those that readers might vaguely recog- nize as potential first names (e.g., Florida, Washington) to those The results of Study 4 strongly suggest that when it comes to people’s first names, implicit egotism is more pronounced among that were extremely obscure (e.g., Arizona, Tennessee). This study women than among men. This finding is consistent with published revealed that people were a full 68% more likely to live in states laboratory research on the name letter effect (Kitayama & Kara- bearing their first names than they should have been based on sawa, 1997; S. L. Koole, 2001, personal communication) as well as chance. Other pilot studies with even more distinctive names with our own recent laboratory research on name letter prefer- yielded even more dramatic results. ences. Women typically like their first names and first initials more If unusual names such as Nevada or Tennessee do an unusually than men do. To the degree that this gender difference proves to be good job of discriminating their owners from every Tom, Dick, replicable, it would strengthen our confidence in the conclusion and Harry they may encounter, then it seems likely that people that name letter preferences per se are responsible for people’s with highly unusual names feel a stronger sense of ownership for attraction to places that resemble their names. In the case of their names. If distinctiveness, like gender, proves to be a robust Study 4, even if one focuses on the female name–state pairs that moderator of our findings, it would further strengthen our inter- bear a less striking resemblance to the states with which they were pretation of our findings. In addition, additional evidence that paired, it appears that our findings were stronger for women than people with distinctive names show stronger than usual name letter for men. Nonetheless, it might be useful to gather additional data preferences would also argue against a mere exposure interpreta- on the moderators of implicit egotism. Along these lines, we tion of our findings. By definition, it is people with highly com-

Table 6 Likelihood of Moving to a Specific Southeastern State as a Function of One’s First Name (Study 4: Results for Women)

First name

State Florence Georgia Louise Virginia Total

Florida 10,062 (8,810) 1,033 5,608 6,706 23,409 Georgia 777 346 (184) 1,118 987 3,228 Louisiana 386 185 506 (402) 467 1,544 Virginia 1,349 336 1,469 2,077 (1,603) 5,231 Total 12,574 1,900 8,701 10,237 33,412

Note. For the diagonal frequencies, in boldface type, expected values (rounded to the nearest whole number) appear in parentheses. 476 PELHAM, MIRENBERG, AND JONES

Table 7 Likelihood of Moving to a Specific Southeastern State as a Function of One’s First Name (Study 4: Results for Men)

First name

State George Kenneth Louis Virgil Total

Georgia 3,592 (3,520) 722 657 204 5,175 Kentucky 2,570 526 (518) 527 299 3,922 Louisiana 2,024 411 699 (476) 118 3,252 Virginia 5,314 964 1,022 198 (309) 7,498 Total 13,500 2,623 2,905 819 19,847

Note. For the diagonal frequencies, in boldface type, expected values (rounded to the nearest whole number) appear in parentheses. mon rather than highly unusual names who are more frequently names listed in column 1. Column 3 lists the proportion of de- exposed to their own names. Study 5 was designed to provide ceased Americans who lived in a given Saint city who possessed further evidence regarding the potential moderators of implicit this same first name. Column 4 lists the (deceased) population for egotism. Study 5 also made use of an even more exhaustive each Saint city. Finally, column 5 lists the effect size (in the form sampling strategy than the one employed in Study 4, as well as an of an odds ratio) for each name–city pair in the list. In interpreting arguably more sophisticated data analytic strategy. The novel these odds ratios, it is important to note that many of the cities analytic approach we adopted in Study 5 allowed us to test our were sufficiently small (or the names sufficiently rare) that the hypotheses for a great number of names without having to generate expected number of residents who possessed that Saint name was an extraordinarily large (e.g., 35 ϫ 35 ϭ 1,225 cell) matrix of less than a single person. For instance, the seemingly troubling names and places of residence. odds ratio of 0 to 1 for women named Agatha living in St. Agatha is not very disappointing. This is because the expected number of Study 5 women named Agatha living in this very small city was less than 1/50th of a person (0.0167 women). City–name combinations that Method yielded expected values of at least 5.0 (meaning that at least 5 In Study 5, we sampled every U.S. city whose name begins with Saint people with that first name should have resided in that specific followed by a person name (e.g., St. Anne, St. Helen, St. Louis, St. Paul). city, based on base rates) are printed in bold. Ten of these 15 odds There are 35 such city names. Because of duplicate city names in different ratios were greater than 1.0. What, then, were the overall results states, however, there are about 80 individual cities with such names. for first names in Study 5? Because we searched for complete first or last names (rather than letter Beginning with the results for women, expected values dictated strings) in Study 5, we made use of the same SSDI database used in that 308.8 of the 45,908 women sampled should have resided in Studies 1 and 4. As noted earlier, SSDI records are extremely comprehensive. Because this search engine places no limit on the number of hits it cities named after Saints who happened to share their first names. returns, it also allows for national base-rate calculations of the frequency of The actual number of women who showed this name–city match- different specific first or last names. Thus, for instance, one can use this ing effect was 445, which is 44% greater than the chance value, tool to determine the exact proportion of deceased Americans whose first ␹2(1) ϭ 41.97, p Ͻ .001. The results for men told a weaker but or last name was Louis. This search engine also allows one to specify the highly significant version of this same story. On the basis of city in which a person was living at the time of his or her death. Thus, for expected values, 3,476.0 out of 594,305 men should have lived in instance, one can determine the exact proportion of deceased residents of Saint cities bearing their first names. The actual number of men St. Louis whose first or last name was Louis. who did so was 3,956, which is 14% greater than the chance value. In Study 5, we exhaustively sampled all of the “Saint” cities in the U.S. and compared (a) the proportion of deceased residents from each city Because of the extremely large sample size for men, this value was 2 whose first name was the same as that city name with (b) the proportion of also highly significant, ␹ (1) ϭ 58.63, p Ͻ .001. A test comparing deceased Americans who possessed that same first name. For example, we the women’s name–city matching rate (144% of the expected were able to determine whether the deceased residents of St. Louis were value) against the name–city matching rate observed for the men disproportionately likely to be named Louis (relative to all other Ameri- (114%) indicated that the gender difference favoring women was cans). We repeated this analysis for all 35 of the Saint city names. In the highly significant, ␹2(1) ϭ 25.1, p Ͻ .001. In short, averaging first such analysis, we searched for first names only (regardless of sur- across the residents of all of these cities, both women and men name). In the second such analysis we searched for last names only gravitated toward Saint cities whose names included their own first (regardless of first name). names, and this tendency was particularly pronounced among women. These findings greatly increase our confidence in the Results and Discussion gender differences observed in Study 4. The entire list of names used in Study 5 appears in column 1 of Presumably, the fact that these city names paired people’s exact Table 8. Column 2 of Table 8 lists the proportion of deceased first names with the positive word Saint may have contributed to Americans in the SSDI database who possessed each of the first their appeal. At the same time, these findings might be especially IMPLICIT EGOTISM IN MAJOR LIFE DECISIONS 477

Table 8 Saint (first) name after whom their hometown was named. The Likelihood of Living in a “Saint” City as a Function of One’s actual number of matches was 128, which is 55% greater than the First Name (Study 5) chance standard, ␹2(1) ϭ 24.97, p Ͻ .001. We also conducted a supplemental analysis of all people whose Proportion Proportion City surname included the word Saint (e.g., Mimi de Saint Aubin, Name U.S. names in city population OR Matthew Saint). We then compared (a) the proportion of people in Women the U.S. whose surname included the word Saint with (b) the 1. Agatha .000091 .0000000 183 0.00 proportion of people living in any Saint city (e.g., St. Paul, St. 2. Anne .001305 .0000000 1,703 0.00 Cloud) whose surname included the word Saint. Whereas there 3. Bernice .001381 .0000000 133 0.00 should have been 242.4 matches on the basis of chance, the actual 4. Clair .000155 .0003153 25,376 2.03 number of matches was 290, ␹2(1) ϭ 9.35, p ϭ .003. Thus, people 5. Helen .009068 .0106762 1,405 1.18 were disproportionately likely to inhabit a Saint city when their 6. Marie .004591 .0059749 5,021 1.30 7. Mary .022972 .0339013 11,504 1.48 surname included the word Saint. Like the primary analysis in- 8. Rose .004141 .0034305 583 0.83 volving surnames, this supplemental finding is not susceptible to any alternate explanations involving either unconscious priming or Men conscious veneration of specific Saints. Like the strong first-name 9. Anthony .002508 .003858 1,296 1.54 findings we observed in the supplement to Study 4 (the study 10. Augustine .000084 .000000 13,057 0.00 involving first names such as Florida and Tennessee), the strength 11. Bernard .001523 .001600 1,250 1.05 12. Charles .014408 .015509 21,343 1.08 of the primary surname matching effect in Study 5 may be 13. David(s) .004549 .002035 2,948 0.45 grounded in the fact that the surnames we sampled were statisti- 14. Elmo .000126 .000000 1,083 0.00 cally rare (in this particular case, rare as surnames). 15. Francis .002432 .004752 2,315 1.95 To gain further insight into whether more distinctive names are 16. Gabriel .000148 .000000 276 0.00 17. George .014347 .012532 6,942 0.87 associated with stronger name letter preferences, we conducted a 18. Henry .006720 .033755 474 5.02 meta-analysis of our individual results for first names. To be sure 19. Ignace .000007 .000000 1,328 0.00 that this analysis was based on reliable estimates of effect sizes for 20. Jacob .001111 .005319 376 4.79 a given name, we limited the analyses to name–city combinations 21. James .020204 .015049 10,499 0.74 22. Joe .002471 .005117 2,345 2.07 that yielded expected frequencies of at least five names (the names 23. John(s) .029861 .022749 5,187 0.76 appearing in bold in Table 8). Because pilot studies suggested that 24. Joseph .013665 .008143 36,349 0.60 the distinctiveness effect might apply primarily to men, and be- 25. Leonard .002038 .002132 469 1.05 cause there were only three female name–city combinations that 26. Louis .004168 .006206 358,699 1.49 27. Mark(s) .000679 .000000 113 0.00 yielded expected values of at least 5.0, we limited this meta- 28. Martin .001477 .000000 77 0.00 analysis to male names. In this analysis, we simply correlated the 29. Matthew(s) .000536 .001037 1,928 1.94 effect size (i.e., the odds ratio) associated with each name with the 30. Michael .003717 .013210 757 3.55 objective frequency of that name (the proportion of Americans 31. Paul .005469 .005445 119,736 1.00 having that name). This analysis indicated that the tendency for 32. Peter .002414 .002956 2,706 1.22 33. Stephen(s) .001221 .000549 1,823 0.45 people to live in cities resembling their first names was notably 34. Thomas .007796 .013746 873 1.76 stronger for the least common first names, r(10) ϭϪ.545, p ϭ 35. Vincent .001080 .000000 56 0.00 .034 (one-tailed). An alternate approach to this analysis would be to categorize each individual participant according to the fre- Note. Names in boldface type have expected frequencies greater than 5.0. Proportion U.S. names ϭ proportion of deceased Americans in the Social quency of his or her name and then examine whether the city– Security Death Index who possessed the first name; Proportion in city ϭ name matching effect was stronger for less common names. It proportion of deceased persons who lived in that particular “Saint” city and should go without saying that such an analysis would yield a ϭ possessed that first name; OR odds ratio. dramatically lower p value because the number of observations for that analysis would be the total sample size associated with prone to an alternate explanation based on priming (or a conscious these 12 names rather than the number 12 (for 12 names). analogue thereof). In particular, the same parents who adore Saint The findings of Study 5 add some punch to the idea that Mary or Saint Louis enough to move a city bearing that Saint’s name-letter preferences influence people’s choice of a place in name might also name one of their sons or daughters after the same which to live. These findings further suggest that when it comes to beloved Saint. Of course, it is not clear why such a priming effect first names, name-letter preferences (a) are stronger for women should hold more strongly for female than for male names. None- than for men and (b) are stronger for unique as opposed to highly theless, it would be reassuring to see that the results of Study 5 common names. Each of these findings increases our confidence in generalized to surnames as well as first names. the conclusion that our findings truly reflect implicit egotism (in As it turns out, all 35 of these Saint names exist as surnames as the form of behavioral name letter preferences). Taken together, well as first names (though in much lower frequencies as sur- these supplemental findings involving gender and distinctiveness names). Were people disproportionately likely to inhabit Saint also suggest that there is more to name letter preferences than mere cities when their surnames were the same as the names appearing exposure. Finally, the robust surname findings for Study 5 further in the city names? Yes. Base-rate calculations indicated that there suggest that the name letter preferences we observed are not should have been 82.6 residents whose surnames matched the merely priming effects in disguise. 478 PELHAM, MIRENBERG, AND JONES

Another Systematic Replication sequences of implicit egotism to an alternate self-evaluative bias. From a methodological perspective, examining whether people’s At the risk of boring some readers, it is worth noting that we birthdays influence major life decisions constitutes an ideal natural have conducted several other systematic replications of our name– experiment on implicit egotism. Because the exact date of a residence matching effects. For example, in a recent replication, person’s birth is essentially a random outcome, this date should be we made use of an exhaustive national telephone directory (the uncorrelated with individual differences such as age, gender, and U.S. Telephone & Address Listings Page at Ancestry.com) to ethnicity. focus on exact surname–city-name matches that should be difficult to explain in terms of any obvious confounds. In particular, we consulted census records to identify the five most common sur- Study 6 names that happen to be place names commonly used in city Method names. These surnames, in order of frequency, were Hill, Park, Beach, Lake, and Rock. People were disproportionately likely to Study 6 made use of the same SSDI records examined in Studies 1, 4, live in cities featuring place words that happened to be their and 5. These records were ideal for the purposes of Study 6 because they surnames. The observed number of matches in this study (1,351) allow searches based on (a) decedents’ exact date of birth as well as (b) the exceeded the chance value (976) by more than 38%, p Ͻ .001. specific city in which decedents lived at the time of their deaths. Further- To summarize thus far, Studies 1–5 strongly suggest that people more, this search engine allows comprehensive searches that include one- are attracted to places that resemble their names. We have sug- word portions of a city name. Thus, if one searches for cities named two, gested that these preferences constitute examples of implicit ego- this search engine will return hits from any U.S. city whose name includes the word two (e.g., Two Harbors, Two Oaks). Study 6 focused on all U.S. tism. However, advocates of explicit egotism might take issue with cities whose names began with the numbers 2–8, inclusive. Typical city this explanation. Perhaps people consciously gravitate toward names included Two Harbors, Minnesota; Three Forks, Montana; and Five places that remind them of themselves (a possibility that, in some Points, Alabama. Although there were a few cities whose names contained ways, is more intriguing than the explanation we have offered). numbers higher than the number 8 (there are no cities, incidentally, We believe there is a big difference between being aware of one’s containing the word one), these cities had extremely small populations, behavior (“My name is Paul, and I am moving to St. Paul”) and making them uninformative. Moreover, adding them to the analysis does being aware of the basis for one’s behavior (“I am doing so not change our findings. To simplify our analysis and to make certain that because of the positive associations I have about my name”). our participants would have a preference for only one birthday number, we Nonetheless, the argument that implicit egotism is responsible for focused on participants born on the same numbered day and month (02/02, the findings of Studies 1–5 would be strengthened by evidence that 03/03, 04/04, etc.). Our list of usable birth dates thus ranged from Febru- ary 2 to August 8. To create the data for Study 6, we created a 7 ϫ 7 matrix more subtle self-relevant associations influence where people that included every possible combination of the seven birthdays and the choose to live. In Study 6 we attempted to gather such evidence by seven sets of city names. We filled in the cells for this 7 ϫ 7 matrix by extending our hypotheses to include an alternate implicit prefer- determining the number of people with each birth date who lived in each ence. Specifically, if people possess positive implicit associations set of cities that corresponded to each number. We expected that, on to the numbers in their birthdays (Kitayama & Karasawa, 1997; average, people whose birthday numbers corresponded to the prominent Miller, Downs, & Prentice, 1998), then it might not be too far- number in each city name would be over-represented among the (deceased) fetched to expect people to gravitate toward cities whose names residents of that specific city. prominently feature these birthday numbers. For example, people born on March 3rd (03/03) and June 6th (06/06) might be attracted Results and Discussion to cities such as Three Rivers, Michigan, and Six Mile, South Carolina, respectively. From a theoretical perspective, support for The complete 7 ϫ 7 frequency matrix for Study 6 appears in this prediction would extend our research on the behavioral con- Table 9. As illustrated in Table 9, the observed frequency in each

Table 9 Likelihood of Living in a Specific City as a Function of One’s Birthday (Study 6)

Birthday

City February 2 March 3 April 4 May 5 June 6 July 7 August 8 Total

Two 22 (17.67) 12 22 16 14 9 24 119 Three 27 40 (28.58) 25 22 32 22 30 198 Four 7 6 11 (7.72) 10486 52 Five 4 2 2 3 (2.49) 23 2 18 Six 5 4 2 4 7 (4.04) 51 28 Seven 1 5 3 2 6 3 (2.99) 525 Eight 6 1 7 10 5 8 8 (7.05) 45 Total 72 70 72 67 70 58 76 485

Note. “City” represents the number in a city name that begins with a number (e.g., Two Harbors, Minnesota; Three Oaks, Michigan). For the diagonal frequencies, in boldface type, expected values appear in parentheses. IMPLICIT EGOTISM IN MAJOR LIFE DECISIONS 479 of the seven diagonal cells exceeded the expected value based on minimum of their first three letters with the names of each of these two chance pairings. That is, for every possible birthday–city set occupations. However, we had to relax our three-letter criterion from Law combination, people were disproportionately likely to have lived in to La for all of the female names (and for three of the four male names) a city whose name prominently featured their birthday number (at because there were no names that qualified using the stricter criterion. least by a very small margin). Whereas the overall percentage of The 16 names we generated in this fashion included the female names Denise, Dena, Denice, Denna, Laura, Lauren, Laurie, and Laverne and the chance matches between birth date and city name should have male names Dennis, Denis, Denny, Denver, Lawrence, Larry, Lance, and been 14.5%, the observed percentage of matches was 19.4%, Laurence. We expected that people with names such as Dennis or Denise ␹2 ϭ ϭ which is about 33% more than the chance value, (1) 9.13, p would be overrepresented among dentists, and people with names such as .003. It thus appears that people are attracted to cities whose names Lawrence or Laura would be overrepresented among lawyers. Because the activate their positive associations regarding their birthday num- lawyer search engine produced a great number of false alarms involving bers. Although we obviously do not know how these people would last or middle as opposed to first names, we carefully cleaned the data have explained their decision to live in a specific city, it seems generated by this search engine to limit hits to true first name hits (no such extremely unlikely that many of these people set out on a con- problem occurred for the dentist search engine). Finally, to make manage- scious trek for a city whose name reminded them of their birth- able the task of cleaning the lawyer data, we limited both searches to the days. In light of the nearly random ways in which people come to eight most populous U.S. states (California, Florida, Illinois, Michigan, New York, Ohio, Pennsylvania, and Texas). possess their specific birthdays, it is also difficult to imagine any compelling alternate accounts of these findings.1 Thus far, we have shown that the positive associations people Results and Discussion have about both the letters in their names and the numbers in their The results for women in Study 7 are summarized in the left- birthdays appear to influence a very important life decision, hand portion of Table 10. Relative to female lawyers, female namely where people choose to spend their lives. In the remaining dentists were quite a bit more likely to have names that began with studies of this report, we focus on another important life deci- the letters Den, ␹2(1) ϭ 4.72, p Ͻ .05. The comparable results for sion—what people choose to do for a living. It is worth noting in men are summarized in the right-hand portion of Table 10. Though advance that the universe of possible careers that resemble peo- the results for men were also in the expected direction, they fell ple’s names is dramatically smaller than the universe of possible short of significance, ␹2(1) ϭ 2.14, ns. Although we expected that places of residence that fit this same bill. This fact, combined with women might show more pronounced name-letter preferences than the scarcity of public databases that include information about men, we did not expect men to show such weak effects.2 people’s names and careers, makes archival studies of implicit In light of the small effect sizes observed in Study 7, critics egotism and career choice inherently difficult to conduct. None- might be concerned that these findings were driven by only one or theless, the same logic that dictates that people should prefer to two of the names we sampled or that they were an artifact of a live in places whose names resemble their own names also dictates specific comparison between dentists and lawyers. We conducted that people should be attracted to careers whose names resemble several analyses to address these concerns, and they all supported their own names. the idea that these findings are a product of implicit egotism. For Because we were unable to locate any large databases that example, in one such analysis, we sampled dentists in all 50 U.S. provided records of people’s names and careers (for multiple states and assessed whether dentists were more likely than the careers), our interest in implicit egotism and career choice neces- average American to possess names such as Dennis, Denis, sitated a change in analytic strategy. In the studies that follow, that Denise, or Dena (the two most common male and female names in is, we had to examine one or two careers at a time. We began our our lists). In this supplemental analysis, we compared (a) the assessment of career choices by focusing on whether people’s first number of dentists with each of these four specific names with (b) names predicted whether they were dentists or lawyers. Of course, the number of dentists who had the two European American names few people ever make a specific choice between these two partic- that were most similar in frequency to each of these specific ular careers. However, comparing the proportion of people with names. For example, according to 1990 census records, the names different first names who choose these two different careers allows for a simple and direct test of our hypothesis 1 The name–city matching effect observed in Study 6 occurred in a Study 7 weaker but significant form in an alternate analysis that included participants born on the 2nd through 8th day of any month of the year. This is a Method very weak test of the name–city matching hypothesis because for 11 out of 12 of the participants in this analysis, the month number associated with We searched for dentists and lawyers by consulting the official Web their birthday conflicts with the day number associated with their birthday. pages of the American Dental Association (http://www.ada.org/directory/ 2 Although both of these online directories allowed searches by first dentistsearchform.html) and the American Bar Association (http://lawyers. name only, the lawyer search directory sometimes returned hits that in- martindale.com/aba). These sources proved to be very useful because they cluded the searched-for name when it proved to be a middle name or a provided comprehensive national directories of the official members of surname. When the first name we searched for showed up as a middle these two professional organizations (though they only allowed searches on name, we included the name on our list because we saw no evidence of any a state by state basis). Each of these directories also allowed for searches systematic bias in which of our names were first versus middle names. based solely on people’s first names, and neither directory placed limits on However, we deleted names from our tally when the searched-for name the number of hits they allowed for a specific search. We began this search proved to be a surname because we observed that some first names turned by consulting 1990 census records. Using these records, we attempted to up as surnames more often than others. Our results do not differ meaning- identify the four most common male and female first names that shared a fully if we exclude all hits that involve middle names. 480 PELHAM, MIRENBERG, AND JONES

Table 10 entific publications in the geosciences (e.g., geology, geophysics, geo- Likelihood of Being a Dentist Versus a Lawyer as a Function of chemistry). Unlike most other widely used bibliographic search tools (e.g., One’s First Name (Study 7) PsycINFO), this search tool generates individual researcher names rather than individual citations (papers). This is important because search tools Women Men that generate individual citations will yield many redundant hits. In Study 8, we examined whether people named George or Geoffrey (the two Occupation Den names La names Den names La names most common American first names beginning with Geo) were disproportionately likely to be published in the geosciences. We did not include Dentist 30 (21.4) 64 (72.6) 247 (229.7) 515 (532.3) additional names in the list because there simply are no such usable names. Lawyer 434 (442.6) 1512 (1503.4) 1565 (1582.3) 3685 (3667.7) We consulted 1990 census data to identify the four European American Note. For each cell of the design, expected values appear in parentheses. male first names that were most similar in overall frequency to these two Each name set is collapsed across the four most common names fitting that target names (using an expanded version of the procedure described in the criterion. The complete set of names used in Study 7 were Denise, Dena, supplemental portion of Study 7). The control names for George were Denice, Denna, Laura, Lauren, Laurie, Laverne, Dennis, Denis, Denny, Daniel, Kenneth, Donald, and Mark. The control names for Geoffrey were Denver, Lawrence, Larry, Lance, and Laurence. Pete, Randolph, Jonathon, and Bennie. Because this search tool does not allow searches by first name only, we had to specify a last name for each search. We selected the eight most common U.S. surnames and paired each Jerry, Dennis, and Walter respectively ranked 39th, 40th, and 41st of the eight surnames with each set of five first names. Doing so had the additional advantage of controlling, in part, for potential ethnic confounds. in frequency for male first names. Taken together, the names Jerry In the interest of brevity we collapsed (a) across the two target names and Walter have an average frequency of 0.416%, compared with (George and Geoffrey), (b) across the eight control names, and (c) across a frequency of 0.415% for the name Dennis. Thus, if people named the eight surnames. Dennis are more likely than people named Jerry or Walter to work as dentists, this would suggest that people named Dennis do, in fact, gravitate toward dentistry. This is the case. A nationwide Results and Discussion search focusing on each of these specific first names revealed 482 On the basis of the observed frequencies for the eight control dentists named Dennis, 257 dentists named Walter, and 270 den- names, there should have been 65.5 geoscientists named George or tists named Jerry. The odds ratio corresponding to this effect is Geoffrey in Study 8. The observed number was 93, or about 42% 1.83 to 1, p Ͻ .001. We conducted identical analyses for the names more than the expected value, ␹2(1) ϭ 4.58, p Ͻ .05. To assess the Denis, Denise, and Dena. Although the sample sizes were much generality of this effect, we conducted an identical analysis for the smaller for these remaining names, the odds ratios for these names three most common male first names that began with the letters Ge also supported our hypotheses. These odds ratios were 1.75 to 1 for (recall that we exhausted the entire list of names beginning with Denis (vs. Vance and Jarrod), 2.50 to 1 for Denise (vs. Beverly and Geo). Although the results for these three names (Gerald, Gene, Tammy), and 1.71 to 1 for Dena (vs. Therese and Candy). Inci- and Gerard) were all in the predicted direction, the sample size for dentally, we did not conduct a comparable set of analyses for each of these names was very small. For instance, the base-rate lawyers because (a) the lower reliability of the lawyer search (control name) calculations for Gerard indicated that there should engine for first names would have required an extraordinary have been a total of 1.75 geoscientists named Gerard in this amount of data cleaning (i.e., removing middle and last names database. In reality there were 4. Thus, this name yielded an from the results, one case at a time) and (b) the lawyer search impressive effect size and an equally unimpressive sample size. engine, unlike the dentist search engine, only allows searches on a In light of this problem, we conducted an additional analysis state by state basis. Critics with large amounts of time on their based on first initials rather than first names (while still using the hands are invited to conduct these analyses for themselves. eight most common U.S. surnames). First, we consulted 1990 The results of Study 7 thus provide additional evidence for the census data for male first names to determine which letter (i.e., idea that implicit egotism plays a role in major life decisions. which initial) provided the closest frequency match for the letter G. Although the methods used to address this question in Study 7 This proved to be the letter T. In 1990, about 8.2% of all American differed slightly from those used in Studies 1–6, observant readers men had first names that began with G, and about 8.9% had first may have noticed that the supplemental analyses presented in names that began with T. Because the large majority of the Study 7 are extremely similar to the primary analyses introduced in scientists listed in the GeoRef index are male, the census estimates Study 5 (the “Saints” study). One desirable feature of this approach for male names probably provide the best estimate of the propor- is that it can be used to make inferences about people’s career tion of geoscientists who should be expected to have first names choices even when one only has access to the names of people that begin with G versus T. Nonetheless, we examined census data working in a single occupation. In Study 8, we made use of this for female first names as well. The percentage of female names approach to see if people whose first names began with the letters beginning with G was 1.8%, and the percentage beginning with T Geo would be disproportionately represented among published was 1.4%. As a stringent indicator of expected frequencies for the professional geoscientists. initials G and T, we thus weighted the male and female percentages equally. On the basis of this estimate, 5.07% and 5.24% of all Study 8 American first names should begin with G and T, respectively. To be certain that these same values held for adults with the Method same eight highly common surnames that were included in Study 8 was conducted using the bibliographic search tool GeoRef Study 9, we consulted the same WorldPages telephone directory (http://georef.cos.com), which contains highly exhaustive records of sci- we used in Studies 2 and 3. We then randomly sampled three U.S. IMPLICIT EGOTISM IN MAJOR LIFE DECISIONS 481 states (Ohio, Massachusetts, and Minnesota), searching each state could then be conducted to identify all of the roofing companies in the for people with each of the eight surnames and each of the two same city. We searched for hardware stores and roofing companies in initials (G and T). This search yielded totals that corresponded the 20 largest U.S. cities (on the basis of 1990 census data). The names of almost perfectly with the estimates based on census data. Specif- these cities appear in Table 11. Because we did not anticipate a large ically, collapsing across these three states (and the eight sur- number of hits for these searches, we collapsed first and last initials into a single analysis and pooled our results across the 20 cities. We searched names), there were 12,510 people whose first initial was G only under hits that appeared under H and R in the alphabetical lists, and and 12,995 people whose first initial was T. It thus seems very safe we were careful to exclude potential hits that appeared to be the names of to assume that among Americans with the eight common surnames streets, towns, or suburbs (e.g., Hollywood Hardware). Examples of hits sampled in Study 9, G is a slightly less common first initial than that we treated as valid indicators of names include HL Campbell Hard- is T. We thus averaged these two estimates (census and phone ware, Harris Hardware, Rashid’s Roofing,H&TRoofing, and Ray & book records) to determine our expected values for geoscientists Roy’s Roofing (counted once). Examples of hits that we excluded from the having these two initials. Relative to the base rates for Americans search include Hardware Mart, Highland Hardware (on Highland Avenue), in general, these geoscientists were much more likely to have first and Roof Roofing (because a phone call to the owner indicated that his names that began with G as opposed to T (at a rate of 177 to 128), company was named after his barking dog). We coded ambiguous hits ␹2(1) ϭ 9.85, p Ͻ .001. Because the standard of comparison in (Hyd-Mech Hardware, Hec Roofing) as half a hit and rounded downward to the nearest whole number in each city (thus it took two ambiguous hits Study 8 was people with exactly the same surnames as these in the same category in a given city to be counted). There were very few geoscientists, it is more difficult than it would be otherwise to of these ambiguous hits, and our results do not change if we exclude them attribute this finding to an ethnic confound. In addition, because from the analysis. Finally, because almost all of these businesses appeared this effect involved people’s first initials rather than the first three to be owned by men, we did not code or analyze for gender. letters in people’s names, it seems unlikely that these people would have realized the connection between their names and their ca- Results and Discussion reers. In short, no matter how we analyzed these data, we found that people whose first names begin with G, Ge,orGeo gravitated As we expected, people’s initials were good predictors of toward the geosciences. whether they worked in roofing or hardware. As shown in the Although the results of Study 8 strengthen our confidence in the upper half of Table 12, hardware store owners were about 80% findings of Study 7, an obvious limitation of Studies 7 and 8 is more likely to have names beginning with the letter H as compared that, collectively, they focus on only three specific career choices. with R. In contrast, roofers showed the reverse pattern. They were Do people’s names have implications for other kinds of careers? In about 70% more likely to have names beginning with R as com- an effort to address this question, we conducted Study 9. In this pared with H. This association between people’s initials and their study, we examined whether people’s first or last initials were occupations was highly significant, ␹2(1) ϭ 10.58, p Ͻ .001. associated with whether they worked in the hardware business or Because we tabulated these data on a city-by-city basis, it was easy the roofing business. to determine whether the roofers in a specific city were more likely than the hardware store owners in that same city to have names Study 9 that began with the letter R. This was the case in 17 of the 20 cities. Although the results of Study 9 supported our predictions, an Method alternate explanation for these results is based on the fact that we In Study 9 we identified owners of hardware stores versus roofing were not able to search for these two businesses by the actual companies whose first or last names began with the letters H versus R.We names of the business owners. Instead, we searched businesses did so by using Yahoo’s Internet Yellow Pages (http://www.yp.yahoo. whose names began with H or R and then examined these hits to com), a search directory that yields alphabetically sorted listings for spe- identify those that involved people’s names. It is thus plausible that cific businesses or companies in specific U.S. cities. For example, a single an equal number of people with the initials H and R go into search using this directory could be conducted to produce an alphabetical hardware and roofing respectively. Instead of reflecting an attrac- listing of all of the hardware stores in Buffalo, New York. A second search tion to careers that resemble one’s name, the disproportionate number of matches we observed between people’s initials and their businesses could merely reflect people’s preference for alliteration. Table 11 For example, if Hector Ramirez opened up both a hardware store The Twenty Largest U.S. Cities in 1990 (the Search Pool and a roofing company, it is plausible that he might choose to for Study 9) name his two businesses Hector’s Hardware and Ramirez’s Roof- ing. In short, people’s love of alliteration rather than their love for Rank City Rank City themselves might be the real reason behind these findings. Of 1. New York, NY 11. San Jose, CA course, if people have a powerful preference for alliteration, they 2. Los Angeles, CA 12. Baltimore, MD should take advantage of other opportunities to provide their 3. Chicago, IL 13. Indianapolis, IN places of employment with highly alliterative labels. We tested this 4. Houston, TX 14. San Francisco, CA alternate hypothesis by conducting three supplemental analyses to 5. Philadelphia, PA 15. Jacksonville, FL 6. San Diego, CA 16. Columbus, OH see if people gravitated toward alliteration in naming their busi- 7. Detroit, MI 17. Milwaukee, WI nesses. We began by searching for hardware stores and roofing 8. Dallas, TX 18. Memphis, TN companies in the eight largest U.S. cities that began with the letters 9. Phoenix, AZ 19. Washington, DC H and R (e.g., Houston, Honolulu, Raleigh, Rochester). We then 10. San Antonio, TX 20. Boston, MA focused on whether people were more likely to name their busi- 482 PELHAM, MIRENBERG, AND JONES nesses after the city in which they lived when it lead to alliteration Table 13 (e.g., “Honolulu Hardware”) than when it did not (“Honolulu Likelihood of Naming a Hardware Store or a Roofing Company Roofing”). Because very few people gave their businesses the Using Non-Self-Referent Words Beginning With H Versus names of the cities in which they lived, we supplemented this R (Study 9) analysis by examining whether the participants in our 20 original cities ever named their businesses after the streets on which they First or last initial were located (e.g., Halsted Hardware on South Halsted Street). Business HR Neither of these analyses yielded any support for the alliteration hypothesis. In fact, the trends in both analyses were weakly in a Hardware 27 (23.85) 22 (22.15) direction opposite that predicted by the alliteration hypothesis. Roofing 32 (32.15) 33 (29.85) Finally, as a third test of the alliteration hypothesis, we recoded the Note. Expected frequencies appear in parentheses. business names from our original 20-city search pool by focusing on business names that began with the letters H versus R but did not designate people’s names (e.g., Handyman Hardware, Hilltop census data and identified the four most common female first names that Roofing, Regency Hardware, Rainbow Roofing). The results of began with the letter pairs Ch and Sh. Of course, we skipped names that did this analysis appear in Table 13. Though weakly in the direction not fit our criterion for pronunciation (e.g., Christine). The eight names we predicted by the alliteration hypothesis, these results fell well short generated in this fashion were Cheryl, Charlotte, Charlene, Cheri, Shirley, of significance, ␹2(1) ϭ 0.39, p ϭ .535. In short, a variety of Sharon, Sherry, and Shannon. Thus for instance, we searched for busi- alternate analyses yielded little support for the alliteration hypoth- nesses beginning “Cheryl’sC” as well as for those beginning “Cheryl’sS.” esis. Nonetheless, it would be useful to devise a test of implicit Two coders were trained to code each hit in this set of searches for whether it truly constituted a business that began with the letter S or C. Hits egotism and career choices that more convincingly ruled out this involving adjectives that began with S or C (e.g., Cheryl’s Creative Arts, alternate explanation. To do this, and to provide a replication Sherry’s Superb Diner) were coded as false alarms. However, when a of our career findings that applied to women, we conducted business name following an adjective constituted a true hit (e.g., Char- 3 Study 10. lotte’s Classic Catering, Cheri’s Super Subs) the adjective was ignored, and it was counted as a regular hit. Along similar lines, in cases in which an Study 10 adjective was an inherent part of the product or service being sold (e.g., Charlotte’s Classical Music) this was also counted as a true hit. Finally, Method because the owners of beauty salons often gave their businesses names that began with S (Sherry’s Salon, Shannon’s Snip & Shear) as well as C The design of Study 10 was very similar to the design of Study 9. Like (Cheri’s Clip n Curl, Charlene’s Classy Cuts), entries that appeared to refer Study 9, for example, Study 10 made use of an electronic telephone to beauty salons were also coded as nonhits. Otherwise, supportive findings directory (http://www.switchboard.com/bin/cgidir.dll). This directory might reflect name letter labeling effects rather than name letter career proved to be ideal for Study 10 because it allows searches using any choices (see Mirenberg & Pelham, 2001). Typical examples of business number of consecutive words or letters that begin a business name. For names that were counted as hits included Cafe´, Candles, Catering, Clothes example, a search for “Kathy’sK” might yield informative hits such as Closet, Sewing, Shoes, Snack Shop, and Sweets. Finally, identical hits “Kathy’s Kite Shop” or “Kathy’s Kayak Kompany.” Thus, after (a) sys- produced in the same city or area code were counted only once—to avoid tematically choosing a set of names and (b) adding an apostrophe, an s, and counting chain stores as multiple hits. Coders were kept blind to the names a single follow-up letter to each name, a researcher can log all of the hits of business owners by replacing the names with arbitrary code numbers. returned in such searches and then code the hits for whether they constitute Interrater agreement was high (␣ ϭ .89). Results were averaged across the valid business names of interest. The flexibility of this search tool also two raters, rounded to the nearest whole number, and summed across each makes it possible to control directly rather than indirectly for alliteration. of the four female first names (within each of the two categories of We did so in Study 10 by capitalizing on the fact that some female first business names). names have the same initial vowel sound despite the fact that they begin with different letters. In particular, quite a few female names that begin with the sh sound begin with the letters Ch rather than Sh. As a concrete Results and Discussion example, if Sheryl is more likely than Cheryl to own a seashell shop, it is The results of Study 10 are summarized in Table 14. As shown difficult to crack this fact up to alliteration. In Study 10, we consulted 1990 in the left-hand column of Table 14, about 23% of women whose names began with Ch appeared to own businesses beginning with the letter S. In comparison, about 39% of women whose names Table 12 Likelihood of Owning a Hardware Store or a Roofing Company as a Function of One’s First or Last Initial (Study 9) 3 We replicated the findings of Study 10 in numerous informal pilot studies as well as in two highly systematic studies, one involving first First or last initial names and one involving surnames. For example, in one study involving surnames we arbitrarily chose computer shops and travel agencies as the Business HRtarget businesses and then generated owner names by sampling the four most common American surnames beginning with the letters C (Clark, Hardware 42 (27.56) 26 (31.44) Carter, Campbell, and Collins) and T (Taylor, Thomas, Thompson, and Roofing 45 (50.44) 76 (57.56) Turner). People whose surnames began with C were slightly more likely to Note. Expected frequencies appear in parentheses. Because of the very be in the computer business (e.g., Campbell’s Computer Repair), and small number of women in this sample, we did not break down the results people whose surnames began with T were quite a bit more likely to be in by gender. the travel business (e.g., Thompson’s Travel). IMPLICIT EGOTISM IN MAJOR LIFE DECISIONS 483

Table 14 last names began with B and G were more likely to contribute to Likelihood of Owning a Business Beginning With the Letter C the election funds of Bush and Gore, respectively. Versus S as a Function of One’s First Name (Study 10)

First name How Representative Are the Present Findings? Initial letter of business name Ch names Sh names Taken as a whole, we feel that our findings provide solid evidence for the existence of implicit egotism. In contrast to this C 55 (46.25) 145 (153.75) view, avid fans of random sampling might argue that it is difficult S 16 (24.75) 91 (82.25) to place great confidence in the present findings. From this perspective, researchers who fail to use random sampling have no Note. Expected frequencies appear in parentheses. assurances that their findings apply to the general population of people about whom they would presumably like to make inferences. Although we sampled names systematically in this research, began with Sh appeared to own such businesses. This association we obviously did not sample them randomly. From our perspec- between name and business type was significant, ␹2(1) ϭ 6.17, tive, random sampling is a methodological tool that, like any other p ϭ .013. Thus, even when we eliminated alliteration as an tool, is well suited to some problems and poorly suited to others. alternate account for our findings, we found that the women in Applying this logic to the present research, one would have to Study 10 gravitated toward businesses that began with the same randomly sample an extraordinarily large number of people in the letter as their first names. We realize that there could be unforeseen hopes of happening across a sizable number of people with sur- risks associated with our use of telephone listings. For example, names such as Califano or Texada. However, it is people with just as it is possible that some people who endorse positive precisely such names who provided the ideal tests of our hypoth- statements on self-esteem scales do not possess high self-esteem, eses (see Mook, 1983, for related arguments). Readers who are it is possible that some of these businesses sampled in Study 10 concerned about whether our results apply to people in general were not truly owned by the women whose names were featured in might be happy to learn that when it comes to studies of implicit their titles. In spite of their potential limitations, however, we egotism and interpersonal attraction, we have been able to use truly believe that Studies 9 and 10 provide a useful complement to the exhaustive sampling techniques. For example, in a study of im- more methodologically rigorous Studies 7 and 8. In other words, plicit egotism and marriage, Pelham et al. (2002) identified every we feel that the most parsimonious explanation for all of our woman who gave birth to a baby in the state of Texas in 1926. As studies involving careers is that implicit egotism plays a role in a a group, these women were married to men who shared their wide range of career choices. Nonetheless, it should go without (maiden) surname initials at a rate that exceeded chance pairing by saying that, given the limited amount of data we were able to more than 40%. Moreover, because this effect also held strongly amass regarding career choices, our confidence regarding implicit for people with Latino surnames, this surname-matching effect egotism and career choices is much lower than our confidence cannot be attributed to ethnic matching. We observed this same regarding implicit egotism and people’s choice of a place in which finding in two large samples of rural Southerners (taken from to live. At a minimum, future studies should examine the most entire counties in Georgia and Florida), in a set of exhaustive, likely moderators of these findings. Like our findings regarding statewide California mortality records, and in exhaustive sets of where people live, for example, our findings regarding careers statewide marriage records from both Georgia and Alabama. In should be moderated by gender as well as by the distinctiveness of fact, we have yet to locate a large sample of marriage records in people’s names. which we did not observe this effect. In our larger samples, we also observed a small but highly significant first name matching effect General Discussion that paralleled the much stronger surname matching effect. More- over, unlike the present studies, which focused mainly on contem- The present studies provide some novel insights into the role of porary samples of European Americans, our studies of relationship the self-concept in major life decisions. In particular, the positive, choices yielded support for implicit egotism in a wide variety of and presumably unconscious, associations people have about cultural and ethnic groups (e.g., rural Southerners, Californians, themselves seem to influence at least two major life decisions: Latinos) and across historical periods ranging from the early 1800s where people choose to live and how people choose to make a to the turn of the 21st century. living. In light of these findings, critics of research on implicit Of course, we are not arguing that concerns regarding sampling social cognition would be hard-pressed to argue that implicit (including random sampling) are completely irrelevant to this self-concept processes have no consequences for important deci- work. For example, a polite take on one reviewer’s critique of this sions. Moreover, although our research on implicit egotism is in its article is that even the most well-intentioned researcher could have early stages, we have already uncovered additional evidence for conducted a very large number of archival studies, selectively the consequences of implicit egotism. For example, we recently presenting only the studies that happen to support his or her found that people are attracted to other people whose names hypothesis. Of course, there is nothing to prevent laboratory ex- resemble their own (Pelham, Jones, Mirenberg, & Carvallo, 2002; perimenters or survey researchers from doing similar things in Jones & Pelham, 2001). For example, Jones and Pelham (2001) their research (Rosenthal, 1963). Nonetheless, the relative ease of found that people’s contributions to political election campaigns conducting archival research should probably exacerbate this par- were influenced by the names of Presidential candidates. More ticular concern. There are several reasons that we do not think this specifically, during the 2000 presidential campaign, people whose concern is applicable to the present report. First, we always chose 484 PELHAM, MIRENBERG, AND JONES our sampling techniques on an a priori basis. Second, we sampled when people consciously try to override them (Wegner, 1994; systematically and often exhaustively. Third, we repeatedly repli- Wegner & Bargh, 1998; Wegner, Ansfield, & Pilloff, 1998). cated our primary findings. Fourth, in several studies we observed As intriguing as these recent laboratory findings are, critics support for theoretically derived moderators of our basic findings. could argue that they merely represent experimental curiosities Finally, for what it is worth, we have never observed a meaningful that have no consequences for important daily decisions. It would reversal of our effect. The numerous pilot studies that we do not be comforting to think so because the possibility that unconscious report here consisted of a great deal of supportive and highly processes influence important day-to-day decisions can be as dis- significant (albeit typically less systematic) studies, combined with turbing as it is intriguing. In other words, the idea that much of a few directionally supportive but nonsignificant replications (typ- what we think, feel, and do is unconsciously determined raises ically involving relatively small samples). From a meta-analytic doubts about some of our most cherished assumptions concerning perspective, we feel very confident that we have identified a personal choice and free will. The current findings add fuel to this meaningful and replicable effect. epistemological flame by suggesting that even major life decisions are not immune to the kinds of unconscious biases that have been The Potency of the Self-Concept repeatedly demonstrated in the laboratory. If decisions as important as where we live or whom we fall in love with are under the By suggesting that implicit egotism is an important determinant control of factors as subtle and capricious as whether our parents of what people like, the findings of this report also attest to the named us Denise or Laura, then this suggests that the feelings of pervasiveness and potency of the self-concept. It is well estab- control and personal choice that we experience when we make lished that people’s consciously reported beliefs about themselves major life decisions are partly illusory (Dennett, 1991; Gilbert, influence a wide variety of important behaviors (e.g., how people 1993; Nisbett & Wilson, 1977; Skinner, 1971; Wegner & Bargh, explain their successes and failures, how long people persist on a 1998). difficult task; Anderson, 1984; Bandura, 1982; Taylor & Brown, Of course, the findings of this report only suggest that uncon- 1987). However, we suspect that existing research has only begun scious processes influence major life decisions if we assume that to scratch the surface of how people’s implicit associations about our participants were typically unaware of the basis of the major themselves influence their thoughts, feelings, and behaviors life decisions we examined. Though we obviously have no way of (Greenwald & Banaji, 1995). By suggesting that people’s associ- knowing what people were thinking when they made these major ations about the letters in their own names influence major life life decisions, we think that in the case of the large majority of our decisions, the present results thus suggest the intriguing possibility findings, people were extremely unlikely to have been aware of the that many of the routes through which the self-concept influences basis for their decisions. Statements such as “I became a lawyer behavior may be unconscious or implicit (Spalding & Hardin, because of the positive associations I have about the letter L,” or 1999). “I took a job in Buffalo because my first name starts with B” simply do not appear in the list of culturally accepted truisms for The Potency of the Cognitive Unconscious why people make important life decisions (Nisbett & Wilson, 1977). Although we consider it well established that people some- The findings of this report also attest to the validity and impor- times know exactly why they do what they do (Quattrone, 1985), tance of a growing body of research on implicit social cognition. the research presented herein suggests that there are also times In the past two decades, an increasing number of studies have when people are completely clueless. In support of the idea that begun to suggest that a great deal of self-regulation and social name letter preferences are truly implicit, Koole, Dijksterhuis, and information processing occurs unconsciously (Banaji & Hardin, van Knippenberg (2001) recently found that name-letter prefer- 1996; Bargh, Chen, & Burrows, 1996; Devine, 1989; Dijksterhuis ences disappeared when people were asked to think carefully about & van Knippenberg, 1998; Epstein, 1990, 1994; Fazio, Jackson, their reasons for liking particular letters. Of course, this is one of Dunton, & Williams, 1995; Greenwald & Banaji, 1995; Higgins, the signatures of implicit social cognition. It is often disrupted by Rholes, & Jones, 1977; Kihlstrom, 1987; Pyszczynski et al., 1999; conscious information processing (Greenwald & Banaji, 1995; Wegner, 1994; Wilson, Lindsey, & Schooler, 2000). Some of the Wilson & Schooler, 1991). most intriguing studies of implicit social cognition suggest that Having said this, we hasten to add that making major life people’s behavior is not always under their own control. For decisions on the basis of implicit egotism is not necessarily irra- example, Bargh et al. (1996) found that college students who had tional. If people feel good when they are exposed to the letters that been subliminally primed to think about old people walked more appear in their own names, then there may be an important sense slowly than usual. Along the same lines, Dijksterhuis and van in which it is highly rational for people to succumb to what Nuttin Knippenberg (1998) found that people who had recently been (1985) called “narcissism without gestalt awareness.” Whereas the primed to think about soccer hooligans rather than college profes- mythological Narcissus was so pathologically drawn to his own sors answered fewer questions correctly on an intellectual task. reflection that it led to his downfall, the more subtle pleasures that Priming effects such as these appear to be most pronounced when people derive from exposing themselves to the letters in their own people are unaware of the priming stimulus (either because it was names seem less likely to lead to self-destruction. As Woody Allen presented subliminally or because people have had time to forget once noted, it may be unduly harsh to criticize people for engaging about it; Bargh, 1992; Strack, Schwarz, Bless, Kubler, & Wanke, in acts of self-love. Specifically, Allen noted, “Don’t knock mas- 1993). At the same time, another potent lesson of recent research turbation; it’s sex with someone I love.” To paraphrase Allen, why on implicit social cognition is that unconscious processes some- should we be so critical of the particular form of narcissism times “have their way” in people’s thoughts and behaviors even identified in this research? From this perspective, falling prey to IMPLICIT EGOTISM IN MAJOR LIFE DECISIONS 485 implicit egotism merely consists of being attracted to that which preferences. Moreover, in each of two nonthreatening control reminds us of the one person most of us love most dearly. conditions, there was no association at all between people’s self- There is also some empirical reason to believe that people are esteem and their name letter preferences. If name letter effects likely to be highly satisfied with the judgments and decisions that were simply mere exposure effects in disguise, there is no reason they make on the basis of automatic affective associations. In that they should become more pronounced for people high in particular, Wilson and Schooler (1991) asked some participants to self-esteem under self-threatening conditions. Koole, Smeets, van think carefully about the reasons for decisions that they would Knippenberg, and Dijksterhuis (1999) also found that name letter normally be expected to make on the basis of their intuitions or preferences are sensitive to psychological threat manipulations. “gut feelings.” For example, in one study they asked some partic- Koole et al. found that people’s liking for their initials was in- ipants to think carefully about their reasons for liking or disliking creased when people were allowed to engage in self-affirmation a variety of different jams. Wilson and Schooler found that intro- after a failure manipulation (presumably, being allowed to self- spection about reasons typically compromised the quality of peo- affirm is the rough experimental equivalent of being chronically ple’s decisions. In one study, jam tasters who thought carefully high in self-esteem). Like Jones et al.’s findings, Koole et al.’s about the reasons for their preferences agreed less than usual with findings suggest that there is more to the name letter effect than expert jam tasters. In another study, students who thought carefully mere exposure. Although we think it is likely that mere exposure about reasons before choosing specific classes were later less plays some role in people’s name letter preferences, we believe satisfied with these classes. These findings suggest that it would be that implicit egotism provides the most parsimonious explanation unwise to equate decisions based on implicit associations with of the total body of research on the name letter effect. irrational or ineffective judgments. If decisions make people feel good in the long run, then it would be best to regard these decisions as rational. Other Likely Moderators of Implicit Egotism

In addition to gender and the distinctiveness of people’s names, Implicit Egotism or Mere Exposure? it seems likely that several other important social and cognitive Just as we have assumed that the findings of this report reflect factors might moderate the strength of people’s behavioral name implicit processes, we have also assumed that they constitute a letter preferences. One likely moderator is suggested by research form of egotism—that is, a bias based on people’s positive asso- on self-verification (Swann, 1987, 1996; Swann, Pelham, & Krull, ciations about the self. To our knowledge, there is only one 1988). Self-verification theory is grounded in the assumption that competing theory that might seem to account for the findings in some people harbor well-developed negative thoughts and associ- this report. In particular, just as Zajonc’s (1968) mere exposure ations about the self. From this perspective, the reason that we effect is loosely consistent with many of Nuttin’s original labora- observed supportive findings in this research is that the large tory findings, it is also loosely consistent with the present findings. majority of people happen to possess positive associations about There can be little doubt, that is, that people are disproportionately themselves. Thus, most people should be attracted to people, exposed to the letters that appear in their own names. At the same places, and things that remind them of themselves. However, when time, given the enormous amount of exposure that fluent readers it comes to the minority of people who possess truly negative have to all the letters in the alphabet, it seems clear that most associations about themselves, it is conceivable that these people people are exposed to highly common letters of the alphabet (e.g., might actually steer away from stimuli that remind them of them- E, S) much more often than they are exposed to the letters in their selves (but cf. Swann, Hixon, Stein-Seroussi, & Gilbert, 1990). own specific names. Nonetheless, the preference that most people Clearly, this would be an interesting direction for future research. have for the letters that appear in their own names appears to be In addition, even if one assumes that all people possess predom- quite a bit more pronounced than the preference that most people inantly positive associations about themselves, this does not guar- have for the most common letters in their own alphabets (Nuttin, antee that people will always gravitate toward people, places, and 1987; Hoorens & Nuttin, 1993). In addition, and in keeping with things that resemble their names. After all, the people, places, and the findings of Kitayama and Karasawa (1997), our findings sug- things to which people could conceivably gravitate often have an gested that when it came to people’s first names, name letter inherent (or conditioned) valence of their own. A great deal of effects were stronger for women than for men (see also Mirenberg research in implicit social cognition (e.g., Fazio et al., 1987, 1995; & Pelham, 2001, who found that this gender difference is reversed Greenwald, McGhee, & Schwartz, 1998) suggests, albeit indi- for people’s surnames). This finding, too, suggests that there is rectly, that people who have positive associations about the self more to the name letter effect than mere exposure. Of course, our might be less attracted than usual to negative stimuli that happen finding that implicit egotism is stronger (at least among men) for to resemble their names. We thus suspect, for example, that be- people with less common names also suggests that there is more to havioral name letter preferences would be stronger than usual for implicit egotism than mere exposure. especially desirable places in which to live, for more prestigious Finally, more definitive evidence that name letter effects may careers, and for more attractive or successful people. Hypotheses constitute a form of self-enhancement comes from a recent study such as these await future scrutiny. As a hypothetical example, a by Jones, Pelham, Mirenberg, and Hetts (2002). Jones et al. as- person named Kruglinski should be especially interested in the sessed participants’ global self-esteem and then exposed some writings of a famous philosopher named Kierkegaard. However, participants to a self-concept threat (writing about a personal flaw). unless this person happened to possess unusually conservative After being exposed to this threat, participants who were high in political views, we would not expect this person to be particularly global self-esteem displayed particularly pronounced name letter attracted to the philosophy of the Ku Klux Klan. 486 PELHAM, MIRENBERG, AND JONES

Summary Fazio, R. H., Sanbonmatsu, D. M., Powell, M. C., & Kardes, F. R. (1986). On the automatic activation of attitudes. Journal of Personality and The findings of this report stand in sharp contrast to many of the Social Psychology, 50, 229–238. assumptions that both scientists and lay people have typically Finch, J. F., & Cialdini, R. B. (1989). Another indirect tactic of (self-) made about major life decisions. For example, these findings raise image management: Boosting. Personality and Social Psychology Bul- serious questions about whether people are fully in control of their letin, 15, 222–232. own behavior. Nonetheless, the idea that people make major life Gilbert, D. T. (1993). The assent of man: Mental representation and the decisions on the basis of unconscious decision rules does not control of belief. In D. M. Wegner & J. W. Pennebaker (Eds.), Hand- necessarily mean that people are irrational. Instead, the specific book of mental control. Century psychology series (pp. 57–87). Engle- form of implicit egotism identified in this research may represent wood Cliffs, NJ: Prentice-Hall. Greenberg, J., & Pyszczynski, T. (1985). Compensatory self-inflation: A an unconscious route through which people create social worlds response to the threat to self-regard of public failure. Journal of Per- that typically make them feel good. Such speculations aside, the sonality and Social Psychology, 49, 273–280. most important implications of these studies may be the most Greenwald, A. G., & Banaji, M. R. (1995). Implicit social cognition: obvious: there may be much more in a name than most people Attitudes, self-esteem, and stereotypes. Psychological Review, 102, realize. To paraphrase an anonymous author of tongue twisters, 4–27. this research offers some new insights into why some people might Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuring find it more satisfying than others to sell seashells by the seashore. individual differences in implicit cognition: The implicit association test. Journal of Personality and Social Psychology, 74, 1464–1480. Hetts, J. J., Sakuma, M., & Pelham, B. W. (1999). Two roads to positive References regard: Implicit and explicit self-evaluation and culture. Journal of Allport, G. W. (1961). Pattern and growth in personality. New York: Holt, Experimental Social Psychology, 35, 512–559. Rinehart & Winston. Higgins, E. T., Rholes, W. S., & Jones, C. R. (1977). Category accessibility Anderson, C. A. (1984). Motivational and performance deficits in inter- and impression formation. Journal of Experimental Social Psychol- personal settings: The effect of attributional style. Journal of Personality ogy, 13, 141–154. and Social Psychology, 45, 1136–1147. Hoorens, V., & Nuttin, J. M. (1993). Overvaluation of own attributes: Mere Banaji, M., & Greenwald, A. G. (1994). Implicit stereotyping and preju- ownership or subjective frequency? Social Cognition, 11, 177–200. dice. In M. P. Zanna, J. M. Olson, et al. (Eds.), The psychology of Huston, T. L., & Levinger, G. (1978). Interpersonal attraction and rela- prejudice: The Ontario symposium, Vol. 7. Ontario symposium on per- tionships. Annual Review of Psychology, 29, 115–156. sonality and social psychology (pp. 55–76). Hillsdale, NJ: Erlbaum. James, W. (1950). The principles of psychology (Vols. 1 and 2). New York: Banaji, M. R., & Hardin, C. D. (1996). Automatic stereotyping. Psycho- Dover. (Original work published 1890). logical Science, 7, 136–141. Jones, J. T., & Pelham, B. W. (2001). Really soft money: The implicit Bandura, A. (1982). Self-efficacy mechanism in human agency. American egotism of political contributions. Manuscript in preparation. Psychologist, 37, 122–147. Jones, J. T., Pelham, B. W., Mirenberg, M. C., & Hetts, J. J. (2002). Bargh, J. A. (1992). Does subliminality matter to social psychology? Name-letter preferences are not merely mere exposure: Implicit egotism Awareness of the stimulus versus awareness of its influence. In R. F. as self-regulation. Journal of Experimental Social Psychology, 38, 170– Bornstein, T. S. Pittman, & S. Thane (Eds.), Perception without aware- 177. ness: Cognitive, clinical, and social perspectives (pp. 236–255). New Kahneman, D., Knetsch, J. L., & Thaler, R. H. (1990). Fairness and the York: Guilford Press. assumptions of economics. In R. Hogarth & M. W. Reder (Eds.), Bargh, J. A., Chen, M., & Burrows, L. (1996). Automaticity of social Rational choice: The contrast between economics and psychology (pp. behavior: Direct effects of trait construct and stereotype activation on 101–116). Chicago: University of Chicago Press. action. Journal of Personality and Social Psychology, 71, 230–244. Kihlstrom, J. F. (1987, September 18). The cognitive unconscious. Science, Beggan, J. K. (1992). On the social nature of nonsocial perception: The 237, 1445–1452. mere ownership effect. Journal of Personality and Social Psychol- Kitayama, S., & Karasawa, M. (1997). Implicit self-esteem in Japan: Name ogy, 62, 229–237. letters and birthday numbers. Personality and Social Psychology Bulle- Crocker, J., & Major, B. (1989). Social stigma and self-esteem: The tin, 23, 736–742. self-protective properties of stigma. Psychological Review, 96, 608– Koole, S. L., Dijksterhuis, A., & van Knippenberg, A. (2001). What’sina 630. name: Implicit self-esteem and the automatic self. Journal of Personality Dennett, D. C. (1991). Consciousness explained. Boston: Little, Brown. and Social Psychology, 80, 669–685. Devine, P. G. (1989). Stereotypes and prejudice: Their automatic and controlled components. Journal of Personality and Social Psychol- Koole, S. L., Smeets, K., van Knippenberg, A., & Dijksterhuis, A. (1999). ogy, 56, 5–18. The cessation of rumination through self-affirmation. Journal of Per- Dijksterhuis, A., & van Knippenberg, A. (1998). The relation between sonality and Social Psychology, 77, 111–125. perception and behavior, or how to win a game of Trivial Pursuit. Kruger, J. (1999). Lake Wobegon be gone! The “below-average effect” and Journal of Personality and Social Psychology, 74, 865–877. the egocentric nature of comparative ability judgments. Journal of Epstein, S. (1990). Cognitive–experiential self-theory. In L. A. Pervin Personality and Social Psychology, 77, 221–232. (Ed.), Handbook of personality: Theory and research. (pp. 165–192). Kunda, Z. (1990). The case for motivated reasoning. Psychological Bul- New York: Guilford Press. letin, 108, 480–498. Epstein, S. (1994). Integration of the cognitive and the psychodynamic Lewicki, P. (1985). Nonconscious biasing effects of single instances upon unconscious. American Psychologist, 49, 709–724. subsequent judgments. Journal of Personality and Social Psychol- Fazio, R. H., Jackson, J. R., Dunton, B. C., & Williams, C. J. (1995). ogy, 48, 563–574. Variability in automatic activation as an unobtrusive measure of racial Miller, D. T., Downs, J. S., & Prentice, D. A. (1998). Minimal conditions attitudes: A bona fide pipeline? Journal of Personality and Social for the creation of a unit relationship: The social bond between birth- Psychology, 69, 1013–1027. daymates. European Journal of Social Psychology, 28, 475–481. IMPLICIT EGOTISM IN MAJOR LIFE DECISIONS 487

Miller, D. T., & Ross, M. (1975). Self-serving biases in the attribution of Staats, C. K., & Staats, A. W. (1957). Meaning established by classical causality: Fact or fiction? Psychological Bulletin, 82, 213–225. conditioning. Journal of Experimental Psychology, 54, 74–80. Mirenberg, M. C., & Pelham, B. W. (2001). [The language of self-love: Strack, F., Schwarz, N., Bless, H., Kubler, A., & Wanke, M. (1993). Name-letter biases in linguistic behavior]. Unpublished raw data. Awareness of the influence as a determinant of assimilation versus Mook, D. G. (1983). In defense of external invalidity. American Psychol- contrast. European Journal of Social Psychology, 23, 53–62. ogist, 38, 379–387. Swann, W. B., Jr. (1987). Identity negotiation: Where two roads meet. Murray, S. L., Holmes, J. G., & Griffin, D. W. (1996). The benefits of Journal of Personality and Social Psychology, 53, 1038–1051. positive illusions: Idealization and the construction of satisfaction in Swann, W. B., Jr. (1996). Self-traps: The elusive quest for higher self- close relationships. Journal of Personality and Social Psychology, 70, esteem. New York: Freeman. 79–98. Swann, W. B., Hixon, J. G., & de la Ronde, C. (1992). Embracing the bitter Nisbett, R. E., & Wilson, T. D. (1977). Telling more than we can know: “truth”: Negative self-concepts and marital commitment. Psychological Verbal reports on mental processes. Psychological Review, 84, 231–259. Science, 3, 118–121. Nuttin, J. M. (1985). Narcissism beyond gestalt and awareness: The name Swann, W. B., Jr., Hixon, J. G., Stein-Seroussi, A., & Gilbert, D. T. (1990). letter effect. European Journal of Social Psychology, 15, 353–361. The fleeting gleam of praise: Cognitive processes underlying behavioral Nuttin, J. M. (1987). Affective consequences of mere ownership: The name reactions to self-relevant feedback. Journal of Personality and Social letter effect in twelve European languages. European Journal of Social Psychology, 59, 17–26. Psychology, 17, 381–402. Swann, W. B., Jr., Pelham, B. W., & Krull, D. S. (1989). Agreeable fancy Paulhus, D. L., & Levitt, K. (1987). Desirable responding triggered by or disagreeable truth? Reconciling self-enhancement and self- verification. Journal of Personality and Social Psychology, 57, 782–791. affect: Automatic egotism? Journal of Personality and Social Psychol- Taylor, S. E., & Brown, J. D. (1988). Illusion and well-being: A social ogy, 52, 245–259. psychological perspective on mental health. Psychological Bulletin, 103, Pelham, B. W., & Hetts, J. J. (1999). Implicit and explicit personal and 93–210. social identity: Toward a more complete understanding of the social self. Tesser, A., & Smith, J. (1981). Some effects of task relevance and friend- In T. Tyler & R. Kramer (Eds.), The psychology of the social self. ship on helping: You don’t always help the one you like. Journal of Mahwah, NJ: Erlbaum. Experimental Social Psychology, 16, 582–590. Pelham, B. W., Jones, J. T., Mirenberg, M. C., & Carvallo, M. (2002). Van Boven, L., Dunning, D., & Loewenstein, G. (2000). Egocentric Implicit egotism: Implications for interpersonal attraction. Manuscript empathy gaps between owners and buyers: Misconceptions of the en- submitted for publication. dowment effect. Journal of Personality and Social Psychology, 79, Perdue, C. W., Dovidio, J. F., Gurtman, M. B., & Tyler, R. B. (1990). Us 66–76. and them: Social categorization and the process of intergroup bias. Wegner, D. M. (1994). Ironic processes of mental control. Psychological Journal of Personality and Social Psychology, 59, 475–486. Review, 101, 34–52. Prentice, D. A., & Miller, D. T. (1992). When small effects are impressive. Wegner, D. M., Ansfield, M., & Pilloff, D. (1998). The putt and the Psychological Bulletin, 112, 160–164. pendulum: Ironic effects of the mental control of action. Psychological Pyszczynski, T., Greenberg, J., & Solomon, S. (1997). Why do we need Science, 9, 196–199. what we need? A terror management perspective on the roots of human Wegner, D. M., & Bargh, J. A. (1998). Control and automaticity in social social motivation. Psychological Inquiry, 8, 1–20. life. In D. T. Gilbert, S. T. Fiske, & G. Lindzey (Eds.), The handbook of Pyszczynski, T., Greenberg, J., & Solomon, S. (1999). A dual-process social psychology (4th ed., Vol. I, pp. 446–496). Boston: McGraw-Hill. model of defense against conscious and unconscious death-related Wills, T. A. (1981). Downward comparison principles in social psychol- thoughts: An extension of terror-management theory. Psychological ogy. Psychological Bulletin, 90, 245–271. Review, 106, 835–845. Wilson, T. D., Lindsey, S., & Schooler, T. Y. (2000). A model of dual Quattrone, G. A. (1985). On the congruity between internal states and attitudes. Psychological Review, 107, 101–126. action. Psychological Bulletin, 98, 3–40. Wilson, T. D., & Schooler, J. W. (1991). Thinking too much: Introspection Rosenthal, R., & Fode, K. L. (1963). The effect of experimenter bias on the can reduce the quality of preferences and decisions. Journal of Person- performance of the albino rat. Behavioral Science, 8, 183–189. ality and Social Psychology, 60, 181–192. Skinner, B. F. (1971). Beyond freedom and dignity. New York: Knopf. Zajonc, R. B. (1968). Attitudinal effects of mere exposure. Journal of Spalding, L. R., & Hardin, C. D. (1999). Unconscious unease and self- Personality and Social Psychology, 9(2, Pt. 2). handicapping: Behavioral consequences of individual differences in implicit and explicit self-esteem. Psychological Science, 10, 535–539. Received June 10, 2000 Staats, A. W., & Staats, C. K. (1958). Attitudes established by classical Revision received September 6, 2001 conditioning. Journal of Abnormal Psychology, 57, 37–40. Accepted September 9, 2001 Ⅲ